An EPYC NixOS Build Farm

Re:An EPYC NixOS Build Farm by davidmr on Saturday August 18, 2018 @02:09PM (#17789716)

I wonder if there were NUMA effects at play since the Intel server was dual-socket and the EPYC was not. It's not entirely surprising that a CPU-heavy workload reading from filesystem cache (i.e. memory) could cause a substantial amount of nonlocal reads that the single-socket server wouldn't have to deal with.

That said, that's just the first thing that popped into my head to explain the result. Either way, the results here are extremely encouraging!

[ Parent | on HN ]

Re:An EPYC NixOS Build Farm by wmf on Saturday August 18, 2018 @02:48PM (#17789931)

A "single" EPYC is eight 4-core processors so there are even more NUMA effects there.

[ Parent | on HN ]

Re:An EPYC NixOS Build Farm by davidmr on Saturday August 18, 2018 @03:13PM (#17790066)

We're now well out of my knowledge base, but your comment certainly makes things interesting. If they have similar NUMA latencies (big if!), I guess maybe the edge would go to whichever chip had higher NUMA bandwidth? Just sort of thinking out loud...

Or perhaps it has absolutely nothing to do with memory speeds, but I can't think of what else would have such a dramatic effect. I'd love to see the same test with a single Xeon.

[ Parent | on HN ]

Re:An EPYC NixOS Build Farm by h4b4n3r0 on Saturday August 18, 2018 @02:52PM (#17789964)

I think it’s the other way around: four 8-core processors

[ Parent | on HN ]

Re:An EPYC NixOS Build Farm by blattimwind on Saturday August 18, 2018 @02:59PM (#17790000)

It's four dice with two groups of cores of four each connected by a callosum.

In NUMA-speak, each group is (probably) a node with neighbouring nodes having a lower distance.

[ Parent | on HN ]